Using syllabic Mel cepstrum features and k-nearest neighbors to identify anurans and birds species

Abstract

Developing efficient methods for monitoring and identifying species of birds and anurans in natural environments are an imperative, in order to attend the concern caused by amphibian decline and trends in decreasing bird population sizes. In this work, a prospective solution to contribute to the mentioned problem is presented by an infrastructure implementation designed to deploy applications in disaster relief and environmental monitoring scenarios, and by formulating a novel application based on Mel-frequency cepstrum coefficients (MFCC), principal components analysis (PCA), and k-nearest neighbors (k-NN) that allows identifying species from segmented syllables in recorded audio. A performance evaluation of the implemented set of algorithms is also presented.